Highlights
Central venous catheter (CVC) insertion is a common procedure among hospitalized patients with more than 5 million CVCs placed annually in the United States.1 As peripherally inserted central catheters (PICCs) can be inserted through a cephalic vein or basilic vein in the upper arm with low risk of complications, they are often a preferred option for long-term intravenous access.2
Immediate complications can occur at the time of catheter insertion and include injury to local structures, including possible nerve injury, phlebitis, air embolism, hematoma, induced arrhythmias, catheter malposition, and rare cardiac and vascular complications.3 The use of ultrasound in combination with the known Seldinger technique used for insertion of CVCs has been demonstrated to yield better puncture success rates and lower incidence rates of postoperative complications.4
Catheter tip malposition is common surrounding PICC insertion, often requiring repositioning or even replacement.5 While the exact position for catheter tip placement continues to be controversial, the lower one-third of the superior vena cava (SVC) at the caval-atrial junction (CAJ) is commonly targeted with upper extremity PICC placement.6 Several methods to confirm PICC line positioning have been proposed, and chest radiography is still the gold standard7; however, identification of exact tip location of PICC line on chest radiography at the CAJ may be less accurate with reports as low as 14% positive predictive value,8 which has prompted investigations into other methods to improve identification and accuracy.
Ultrasonography (US), one of the proposed mechanisms for catheter placement confirmation, has several proposed advantages over radiography including the lack of ionizing radiation, decreased cost, and diagnostic time.9 There have been limited studies evaluating the use of bedside US to confirm PICC positioning. We conducted a prospective observational study to evaluate the feasibility of using US for PICC placement positioning confirmation among new US users (pediatric intensive care trainees).
Materials and Methods
Study Design
This study was performed from August 2017 to May 2018 at Primary Children's Hospital (PCH) in Salt Lake City. PCH has a well-established PICC placement team consisting of nurses who place on average 240 PICCs a year. The study protocol was approved by the Institutional Review Board at the University of Utah as a quality improvement study with waiver of informed consent.
All hospitalization patients under the age of 18 years with a planned PICC placement were eligible. Enrollment was dependent on the availability of the study team primary investigators (PIs).
The PICC team routinely uses standard universal precautions with a modified Seldinger technique and ultrasound guidance for PICC placement. The basilic vein or cephalic vein is typically cannulated with the aid of a Sonosite S-series® US and linear probe (FUJIFILM Sonosite Inc, Bothwell, WA) for insertion with a 0.012–0.18-inch guide wire (Galt®; Galt Medical Corp, Garland, TX). The PICC length is predetermined in each patient by measuring the distance from the estimated insertion site to sternal notch with the arm 45–90° away from the body and palm turned up. The catheter is cut to the estimated length plus 15 cm, and a stiffening wire is preloaded. After vein cannulation and guide wire insertion, the PICC team then inserts a micro-introducer or dilator over the guide wire and into the vein, removes the guide wire, and inserts the catheter with the preloaded stiffening wire. After threading the catheter, stiffening wire is removed, and the PICC team injects saline through the newly inserted catheter while scanning the internal jugular (IJ) vein with the US for presence of turbulent flow, indicating the catheter has traveled cephalad.
Study Team
All US studies were performed by 2 pediatric intensive care fellows in their third year of fellowship with extensive experience in procedural US use but minimal formal training in US. One fellow had completed the Society of Critical Care Medicine US training course. Both had used US throughout fellowship to aid with procedures such as CVC, chest tube, and arterial catheter placement. They underwent 1 day of training in proper technique to confirm PICC placement using the GE Vscan dual probe handheld US® (GE Healthcare, Milwaukee, WI). Training education was performed by a member of the National Board of Echocardiography with testamur status and an international expert in critical care bedside ultrasound.
Ultrasound Technique
In this study, once the PICC team cannulated the vessel, the radiograph technician was notified, and the PI performed a bedside US evaluation with 2 standard cardiac views (subcostal and apical). All members of the PICC team and radiology department were blinded to the US results findings. The PIs recorded if they believed the tip was in the correct position, which means they saw the positioning, quality of the view, and how confident they were in the findings. US images were recorded for blind review by an expert in critical care US to give an independent blinded verification of proper or improper catheter positioning.
The study team employed 2 primary echocardiographic views (subcostal 4-chamber view and the apical 4-chamber view) for tip confirmation in the SVC or right atrium (RA). All views were obtained with the GE Vscan dual probe handheld US®, using the cardiac probe. These views were selected for the following reasons: (1) clear pattern recognition of obtaining proper window, (2) ease of teaching and obtaining 1 of the views with a supine patient regardless of intubation, and (3) because the angle would provide a clear dot of the wire in cross-section in the RA. Bedside US catheter positioning confirmation was performed via direct US identification of the PICC wire tip within the RA and by indirect confirmation of tip location by rapid entry of turbulent flow into the RA or surrounding vascular structures after rapid saline flush of the PICC (Figures 1 and 2). Turbulent flow was adequately achieved by standard fast saline flush through the catheter with resulting microbubbles. This technique was based on prior techniques established by infusing agitated saline for echocardiography when identifying septal wall defects. Both rapid flush and agitated microbubbles can be identified on routine cardiac US. While agitated microbubbles are larger and easier to see, rapid flush creates turbulent flow that is also identifiable on US, and the study protocol did not involve any changes to standard PICC line procedures. Agitated saline could have been more reliable because the bubbles are larger and easier to see, but adding this to the study might have introduced an accidental air embolism from improper technique. We therefore used visualization of turbulent flow from standard catheter flushing for the safety of the patients.
The PI attempted both US views on each patient as permitted with the cardiac probe. An attempt was made on each patient to visualize the inferior vena cava (IVC), and the presence or absence of the wire was also recorded. The wire was then removed, and an indirect confirmation test was performed with a rapid saline push. A member of the PICC team would state aloud when the saline was pushed, and the seconds were counted before the turbulent flow was visualized by the study team (saline flush visualization time).
Outcomes Measures
The primary outcomes, determined a priori, were the sensitivity and specificity of US confirmation of PICC tip positioning. Secondary outcomes measured were the feasibility of adequate US image acquisition (subcostal and apical views) with quality of image rated on a scale of 1–3, as well as ability for image acquisition of insertion wire and turbulent flow in the RA. The feasibility of teaching the technique with a similar sensitivity and specificity for adequate tip positioning between the trainees and the expert was also a reported result.
Images were captured, saved, and graded as 1 = fair, 2 = good, or 3 = great. The outcome measure of diagnostic image quality scale was created as a continuous scale with scoring based on prior studies implementing a provider US training curriculum.10 The scale details included ability to identify anatomical landmarks, interpretation of findings after image acquisition, and quality of image details. Fair was defined as minimally acceptable sufficient information to support interpretation. Good was defined as reasonable quality and ease of identification. Great was defined as easy to identify landmarks and quality of image as good as a trained provider would be able to obtain with same capacity of patient demographics and US machine. Images were further reviewed with the same scale by an US-certified physician (EH) blinded to the US operator's read and to the radiographic results.
Balance Measures
The PI also recorded how many chest radiographs were required per patient until final proper position was confirmed, time from radiology technician call until arrival, as well as if the PICC team noted the presence of turbulent flow from the saline flush in the IJ. The confidence level of the performing fellow on their technique and ability to obtain accurate images was recorded on a scale from 1 to 5, with 1 = minimal confidence and 5 = very confident.
Statistical Analysis
Patient demographic data including age, weight, and location of PICC placement were collected. Demographic and clinical characteristics are expressed as median or interquartile range (IQR). For comparison of continuous variables between 2 groups, a Student's t test was used for normally distributed data and a Wilcoxon rank sum test for skewed data. Categorical variables were compared using contingency tables or Fisher's exact test. Reasons for difficulty in image acquisition were recorded for each patient. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy of US were calculated among successful cases and compared to chest radiography. A P value < 0.05 was considered statistically significant.
Results
A total of 28 patients were enrolled during the 10-month study period. The average age of the patients was 10 years and ranged from 0 to 18 years of age with the largest age group between 11 and 18 years old and median weight of 34 kg. The most common location performed was in the special procedure unit (57%; Table 1). Of the 28 patients, a subcostal 4-chamber view was obtained in 96% of those attempted (n = 26, 92%), and an apical 4-chamber view was obtained in 100% of those attempted (n = 9, 32%; Table 2). The 3 patients in whom a subcostal view was not obtained were secondary to an abdominal dressing or abdominal pain. The quality of subcostal view was rated as great 32% of the time, and the quality of the apical view was great 44% of the time (Table 3). The confidence level of the operators performing US over time is shown in Figure 3 and did not increase over the study period. This was true when both operators’ confidence levels were independently evaluated. There was weak correlation with confidence level and age or weight of the patient (r = −0.25 and r = −0.19, respectively).
| n=28 | |
|---|---|
| Agea (y) | 10 (5, 14) |
| 0–1 | 1 (4%) |
| 1–5 | 7 (25%) |
| 6–10 | 9 (32%) |
| 11–18 | 11 (39%) |
| Weighta (kg) | 34 (20, 55) |
| Location of peripherally inserted central catheter placement | Special procedure unit: 16 (57%) |
| Pediatric intensive care unit: 5 (18%) | |
| Floor: 3 (10%) | |
| Cardiac intensive care unit: 2 (7%) | |
| Emergency department: 1 (4%) | |
| Operating room: 1 (4%) | |
| Catheter length (cm)a | 29 (23, 34) |
a Mean and standard deviation
| Number visualized with ultrasound attempted, n (%) | Total number attempted, n (%) | |
|---|---|---|
| Subcostal | 25 (96) | 26 (92) |
| Apical | 9 (100) | 9 (32) |
| Inferior vena cava | 15 (54) | 28 (100) |
| Wire | 7 (25) | 28 (100) |
| Central turbulent flow seen | 24 (100) | 24 (86) |
| ≤2 s | 20 (79) | |
| ≤3 s | 4 (21) |
| Table head | Image quality |
|---|---|
| Subcostal n (%) | (1) 8 (32) |
| (2) 8 (32) | |
| (3) 9 (36) | |
| Apical | (1) 4 (44) |
| (2) 4 (44) | |
| (3) 1 (11) |
a(1) Fair = minimally acceptable sufficient information to support interpretation. (2) Good = reasonable quality and ease of identification. (3) Great = easy to identify landmarks and quality of image as good as a trained provider would be able to obtain with same capacity of patient demographics and ultrasonography machine
The IVC was only visualized 54% of the time (Table 2). Surprisingly, the wire itself was only visualized 25% of the time. Three of these were seen deep in the RA or right ventricle, and only 4 wires were ever seen at the CAJ in good position.
The indirect measurement of PICC placement with saline flush turbulent flow visualization was attempted in 86% of patients and visualized in 100% of these attempted (Table 2). The only time central turbulent flow (defined as turbulent flow from the saline push visualized in the heart) was not seen was when the PICC was placed in the IJ (identified with Sonosite US by the PICC team) or deep in the right ventricle past the tricuspid valve (location identified on chest radiograph). The median saline flush visualization time when the PICC was later confirmed to be in appropriate positioning was 1.5 seconds. In 6 cases, the line was identified on chest radiograph as malpositioned (deep in RV = 4 and above the CAJ = 2), but the saline flush visualization time was still ≤2 seconds. However, in all other malpositioned lines (4) where turbulent flow was still seen after saline flush, the saline flush visualization time was delayed to a median of ≥3 seconds.
The mean number of chest radiographs per patient was 1.5 (IQR 1–3) with 50% of patients requiring more than 1 image. The mean time spent waiting on chest radiographs was 7 minutes, and this did not vary based on location PICC was placed (P = 0.57). PICCs were identified as deep on official chest radiograph in 6 patients (Table 4). This was identified on US 3 times, giving a PPV of 43% and NPV of 86% for the US identifying deep lines (Table 5). The overall PPV of US identifying malpositioned lines in this study was 43%.
| N=28 | |
|---|---|
| Number of chest radiographs per patient | 1 (1, 2.5)a |
| Time awaiting first chest radiograph | 7 (5, 10)a |
| PICC identified on CXR at cavo-atrial junction | 17 (60)b |
| PICC identified as deep on chest radiograph | 6 (21)b |
| PICC identified in the neck | 2 (7)b |
| PICC identified other locations (innominate, subclavian) | 3 (11)b |
CXR = chest radiograph; PICC = peripherally inserted central catheter.
aMedian and interquartile range.
bN and percentage
| Sensitivity | Specificity | PPV | NPV | |
|---|---|---|---|---|
| Ultrasound identified malpositioned lines | 27 | 76 | 43 | 62 |
| Ultrasound identified deep line | 50 | 81 | 43 | 86 |
NPV = negative predictive value; PPV = positive predictive value
When the US trained physician reviewed the saved images, 92% of images were available for review (Table 6). Images on 2 patients had not been properly saved. Identification of turbulent flow was similarly seen between reviewer and study trainee operators 54% of the time, and quality of image rating was ranked similarly 62% of the time. The wire was visualized 6 out of the 7 times the PI identified it (86%), with 33% when the line was deep.
| n=26 (92%) | |
|---|---|
| Images available to review | 136 |
| Wire visualized | |
| Yes | 6 (26%) |
| No | 17 (74%) |
| Rating of image quality | (1) 2 (9%) |
| (2) 9 (39%) | |
| (3) 12 (52%) | |
| Central turbulent flow visualized | 13 (57%) |
a(1) Fair = minimally acceptable sufficient information to support interpretation. (2) Good = reasonable quality and ease of identification. (3) Great = easy to identify landmarks and quality of image as good as a trained provider would be able to obtain with same capacity of patient demographics and ultrasonography machine
Discussion
Our study did not demonstrate that a bedside US performed by a trainee can safely replace the current gold standard of chest radiography. We found that a subcostal view may be quickly and easily obtained during standard PICC placement procedures in most patients unless physical obstruction to the view is present such as an abdominal dressing or wound vacuum. Despite the ability to accurately obtain the subcostal 4-chamber view, and the inferior CAJ, proper wire position was only visualized less than one-quarter of the time. Our US confirmation protocol performed more favorably when identifying a deep catheter with a sensitivity and specificity of 51% and 81%. Currently, a chest radiograph is performed with each adjustment of the line positioning, and each radiograph requires the placement of the machine and removal of staff from the area of radiation exposure. By using US to identify deep placement, the line may be adjusted in real time. Although our US confirmation protocol of PICC line placement cannot replace the gold standard radiograph, it is possible that a PICC line placement protocol that expands the use of US to additional echocardiographic windows and views may have better sensitivity and specificity and could ultimately reduce the overall procedural timing and/or radiation exposure.
We believe there were a number of reasons for our results, including size and echogenicity of PICC introducer wires, procedural technique used by the PICC team, the US device chosen for this study, and novice US training of fellows performing the US. Limited previous reports describe the use of US in PICC placement confirmation, but multiple studies describe the effective use of US for CVC positioning confirmation compared with chest radiographs.11–13 The majority of these studies examine US accuracy in placement of CVCs placed in the IJ vein and subclavian vein in adults. Matsushima et al. looked at all CVCs including PICCs, and of 83 catheters placed, 41 were PICCs.2 They report 5 false negatives, and of the 5 missed malpositioned catheters, 4 were in PICCs. The size of PICCs and the corresponding wire size could affect the ability of US to identify with ease and explain our high NPV rates. Zaghloul et al. looked at neonates and showed high interrater reliability between bedside US and chest radiographs to identity malpositioning of CVC tips14; however, the highest agreement coefficient was found in neonates weighing <1000 g, and they state the advantage in neonates’ clear acoustic windows. A thin echogenic catheter tip could possibly mitigate this difficulty.
When it was recognized that the wire was not easily visualized in the CAJ or RA, roughly 4 months into the study period, we performed an internal check and education to view PICC wires in coordination with the Blue Phantom gel block and the GE Vscan US. Ultimately, we identified that the wire was not easily visualized with the GE Vscan even on the gel block. For methodological fidelity, we continued the protocol without changing the US machine. The GE Vscan US is designed for adults with image depth optimization on the cardiac probe of 8 cm (range 6–12 cm possible). The depth required for most of the pediatric patients included was 6 cm, and we could have benefited from 4 cm. Real-time scanning with this machine limited resolution at shallower depths and proved more challenging than initially anticipated. This challenge was also a factor when the US certified physician reviewed the images and only 52% of the saved images were rated as great in quality.
Another possible reason we were unable to accurately confirm placement of PICCs could be due to the procedural technique of the local PICC team. Their technique involves inserting a 50 cm guide wire and then subsequently a micro-introducer. They then pull back a stiffening wire to the desired length of the previously trimmed catheter and insert together. This is in contrast to many Seldinger CVC insertion techniques for which the wire is advanced into the RA and the catheter placement is not dependent on catheter tip and wire tip alignment. This commonly accepted approach to PICC placement makes visualization of the wire in an appropriate position more difficult unless you are able to precisely obtain images of the SVC atrial junction in its entirety, which was only possible in 14% of our patients.
Despite this technical hurdle, we were able to identify central turbulent flow in less than 2 seconds in 79% of patients. Prior reports have also favored the measurement of “push-to-bubble” time (saline flush visualization time), or a similar name for this procedure, as an accurate way to verify CVC correct positioning. The application of US confirmation of CVC with visualization of saline originates from the use of the echocardiographic “bubble test.” The traditional bubble test has been used by cardiology to identify septal wall defects by agitating a saline flush by hand and injecting the microbubbles through a catheter while performing an echocardiogram to visualize flow of bubbles through the heart.15 The microbubbles created with saline agitation are too large to pass through the pulmonary system, and therefore, if they are visualized on the left side of the heart, a septal defect is present. Prior reports have shown that visualization of turbulent saline in the RA within 2 seconds of flush confirms CVC placement with 96% sensitivity and 93% specificity.16 However, a standard amount of time with PICC lines has not been fully established, and identification of central bubbles did not equate to confirmation of central depth location in this study. Two things are important to note in our study. First, we did not use agitated saline. Central turbulent flow refers to turbulent flow seen during normal flushing of the central line with a normal saline syringe. Second, time to visualization of central turbulent flow in PICC line insertion could represent an easier and more reliable method of confirming PICC line placement with US, but the exact time cutoff would need to be tested in another larger, more specifically designed study.
With US utilization increasing among critical care fellowship training, we had hoped that minimal training in US technique would suffice for this study design. In prior reports, specifically in the pediatric population, supporting the use of US as an accurate option over chest radiograph for CVC positioning, study operators had extensive formal US training.14,17 There was interestingly no increase in confidence over the time period of 10 months among either fellow performing the study. We suggest that more hands-on training specifically with PICC line catheters and more practice in obtaining the CAJ view prior to this study could have been beneficial instead of measuring this as an objective.
Limitations
Our study has several limitations beginning with its small sample size. We enrolled only 28 patients, which took 10 months to complete. We employed a portable handheld GE Vscan ultrasound designed for adult patients to maximize efficiency, but we likely sacrificed image accuracy. The study operator's ability to accurately visualize PICC wire placement during this study was likely influenced significantly by learner effect and the highly operator-dependent nature of US technique and image interpretation. A great deal of variability was identified in turbulent flow visualization between the study group and the US trained physician. This is likely explained by the real-time benefit of knowing when the saline is infused during image acquisition; however, the reverse is also possible, and expecting an image can introduce a visual bias. The study design was quality improvement; therefore, we did not alter PICC procedures or delay insertion time by carefully searching for the catheter position. Identification of turbulent saline flow turned out to be a more reliable indication of catheter tip placement. We did not perform the study in real time with an expert in critical care ultrasound placement performing the diagnostic techniques.
Future Direction and Lessons Learned
We learned several things from this quality improvement study. Confirming PICC placement via US without ease of visualization of the wire proved challenging in that the study trainees were never able to gain confidence of exactly what to look for with proper placement. A repeat study with use of an echogenic catheter tip would likely improve US accuracy and ease of identification. This would however require a change in hospital equipment and procedure. We also feel more extensive and possibly formal training is needed for novice US users to master the skills needed to detect small wires in the heart and SVC. It was very challenging for novice US users to detect a small wire with a pediatric patient's anatomy using a portable US machine with lower resolution and only adult specifications. In hindsight, we would have hoped to have increased accuracy and ease of identification with a larger US machine with better resolution and pediatric specifications.
The negative results we experienced may be more a function of the protocol than a statement about the efficacy of US in confirming PICC line placement. A protocol that visualized the tip deep in the RA and then pulled the wire back until it was no longer visualized may have produced different results, both in diagnostic accuracy and in trainee confidence. A slight modification to the current practice might be necessary for the quality project to have succeeded. A more precise measurement of tenths of a second during saline injection and turbulent flow visualization could have been more accurate in providing a cutoff time for malposition as well.
More frequent review of saved images by the trained physician could have identified difficulty in wire visualization earlier and led to supervised US and possible improvement in confidence. Regardless, this quality improvement study presents a first step in trying to use new noninvasive technology to improve patient safety and reduce wait times or radiation exposure.
Conclusions
Catheter tip malposition is a common complication associated with CVC or PICC insertion. Unfortunately, our limited 2-view US protocol, performed by trainees in ultrasound use, did not favorably compare to confirmation with chest radiograph. There remains, however, potential for other US protocols, with pediatric-specific technology or echogenic catheter tips to reduce radiation exposure from chest radiograph during PICC line positioning verification.